| Message ID | 20210130004401.2528717-3-ndesaulniers@google.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | Kbuild: DWARF v5 support | expand |
On Sat, Jan 30, 2021 at 1:44 AM Nick Desaulniers <ndesaulniers@google.com> wrote: > > DWARF v5 is the latest standard of the DWARF debug info format. > > Feature detection of DWARF5 is onerous, especially given that we've > removed $(AS), so we must query $(CC) for DWARF5 assembler directive > support. > > The DWARF version of a binary can be validated with: > $ llvm-dwarfdump vmlinux | head -n 4 | grep version > or > $ readelf --debug-dump=info vmlinux 2>/dev/null | grep Version > > DWARF5 wins significantly in terms of size when mixed with compression > (CONFIG_DEBUG_INFO_COMPRESSED). > > 363M vmlinux.clang12.dwarf5.compressed > 434M vmlinux.clang12.dwarf4.compressed > 439M vmlinux.clang12.dwarf2.compressed > 457M vmlinux.clang12.dwarf5 > 536M vmlinux.clang12.dwarf4 > 548M vmlinux.clang12.dwarf2 > > 515M vmlinux.gcc10.2.dwarf5.compressed > 599M vmlinux.gcc10.2.dwarf4.compressed > 624M vmlinux.gcc10.2.dwarf2.compressed > 630M vmlinux.gcc10.2.dwarf5 > 765M vmlinux.gcc10.2.dwarf4 > 809M vmlinux.gcc10.2.dwarf2 > > Though the quality of debug info is harder to quantify; size is not a > proxy for quality. > > Jakub notes: > All [GCC] 5.1 - 6.x did was start accepting -gdwarf-5 as experimental > option that enabled some small DWARF subset (initially only a few > DW_LANG_* codes newly added to DWARF5 drafts). Only GCC 7 (released > after DWARF 5 has been finalized) started emitting DWARF5 section > headers and got most of the DWARF5 changes in... > > Version check GCC so that we don't need to worry about the difference in > command line args between GNU readelf and llvm-readelf/llvm-dwarfdump to > validate the DWARF Version in the assembler feature detection script. > > GNU `as` only recently gained support for specifying -gdwarf-5, so when > compiling with Clang but without Clang's integrated assembler > (LLVM_IAS=1 is not set), explicitly add -Wa,-gdwarf-5 to DEBUG_CFLAGS. > > Disabled for now if CONFIG_DEBUG_INFO_BTF is set; pahole doesn't yet > recognize the new additions to the DWARF debug info. Thanks to Sedat for > the report. > > Link: http://www.dwarfstd.org/doc/DWARF5.pdf > Reported-by: Sedat Dilek <sedat.dilek@gmail.com> > Suggested-by: Arvind Sankar <nivedita@alum.mit.edu> > Suggested-by: Caroline Tice <cmtice@google.com> > Suggested-by: Fangrui Song <maskray@google.com> > Suggested-by: Jakub Jelinek <jakub@redhat.com> > Suggested-by: Masahiro Yamada <masahiroy@kernel.org> > Suggested-by: Nathan Chancellor <natechancellor@gmail.com> > Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> > --- > Makefile | 1 + > include/asm-generic/vmlinux.lds.h | 7 ++++++- > lib/Kconfig.debug | 18 ++++++++++++++++++ > scripts/test_dwarf5_support.sh | 8 ++++++++ > 4 files changed, 33 insertions(+), 1 deletion(-) > create mode 100755 scripts/test_dwarf5_support.sh > > diff --git a/Makefile b/Makefile > index d2b4980807e0..5387a6f2f62d 100644 > --- a/Makefile > +++ b/Makefile > @@ -831,6 +831,7 @@ KBUILD_AFLAGS += -Wa,-gdwarf-2 > endif > > dwarf-version-$(CONFIG_DEBUG_INFO_DWARF4) := 4 > +dwarf-version-$(CONFIG_DEBUG_INFO_DWARF5) := 5 > DEBUG_CFLAGS += -gdwarf-$(dwarf-version-y) > > ifdef CONFIG_DEBUG_INFO_REDUCED > diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h > index 34b7e0d2346c..1e7cde4bd3f9 100644 > --- a/include/asm-generic/vmlinux.lds.h > +++ b/include/asm-generic/vmlinux.lds.h > @@ -842,8 +842,13 @@ > /* DWARF 4 */ \ > .debug_types 0 : { *(.debug_types) } \ > /* DWARF 5 */ \ > + .debug_addr 0 : { *(.debug_addr) } \ > + .debug_line_str 0 : { *(.debug_line_str) } \ > + .debug_loclists 0 : { *(.debug_loclists) } \ > .debug_macro 0 : { *(.debug_macro) } \ > - .debug_addr 0 : { *(.debug_addr) } > + .debug_names 0 : { *(.debug_names) } \ > + .debug_rnglists 0 : { *(.debug_rnglists) } \ > + .debug_str_offsets 0 : { *(.debug_str_offsets) } > I just looked at binutils 2.36 in the Debian/experimental repositories. [1] says: + PR ld/27230 + * scripttempl/DWARF.sc: Add DWARF-5 .debug_* sections. ... - /* DWARF Extension. */ - .debug_macro 0 : { *(.debug_macro) } + /* DWARF 5. */ .debug_addr 0 : { *(.debug_addr) } + .debug_line_str 0 : { *(.debug_line_str) } + .debug_loclists 0 : { *(.debug_loclists) } + .debug_macro 0 : { *(.debug_macro) } + .debug_names 0 : { *(.debug_names) } + .debug_rnglists 0 : { *(.debug_rnglists) } + .debug_str_offsets 0 : { *(.debug_str_offsets) } + .debug_sup 0 : { *(.debug_sup) } The list of DWARF-5 .debug_* sections is alphabetically sorted. AFAICS .debug_sup section is missing? - Sedat - [1] https://salsa.debian.org/toolchain-team/binutils/-/commit/f58f33081035672b01a04326a9c8daadbd09a430 > /* Stabs debugging sections. */ > #define STABS_DEBUG \ > diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug > index 94c1a7ed6306..ad6f78989d4f 100644 > --- a/lib/Kconfig.debug > +++ b/lib/Kconfig.debug > @@ -268,6 +268,24 @@ config DEBUG_INFO_DWARF4 > It makes the debug information larger, but it significantly > improves the success of resolving variables in gdb on optimized code. > > +config DEBUG_INFO_DWARF5 > + bool "Generate DWARF Version 5 debuginfo" > + depends on GCC_VERSION >= 50000 || CC_IS_CLANG > + depends on CC_IS_GCC || $(success,$(srctree)/scripts/test_dwarf5_support.sh $(CC) $(CLANG_FLAGS)) > + depends on !DEBUG_INFO_BTF > + help > + Generate DWARF v5 debug info. Requires binutils 2.35.2, gcc 5.0+ (gcc > + 5.0+ accepts the -gdwarf-5 flag but only had partial support for some > + draft features until 7.0), and gdb 8.0+. > + > + Changes to the structure of debug info in Version 5 allow for around > + 15-18% savings in resulting image and debug info section sizes as > + compared to DWARF Version 4. DWARF Version 5 standardizes previous > + extensions such as accelerators for symbol indexing and the format > + for fission (.dwo/.dwp) files. Users may not want to select this > + config if they rely on tooling that has not yet been updated to > + support DWARF Version 5. > + > endchoice # "DWARF version" > > config DEBUG_INFO_BTF > diff --git a/scripts/test_dwarf5_support.sh b/scripts/test_dwarf5_support.sh > new file mode 100755 > index 000000000000..c46e2456b47a > --- /dev/null > +++ b/scripts/test_dwarf5_support.sh > @@ -0,0 +1,8 @@ > +#!/bin/sh > +# SPDX-License-Identifier: GPL-2.0 > + > +# Test that the assembler doesn't need -Wa,-gdwarf-5 when presented with DWARF > +# v5 input, such as `.file 0` and `md5 0x00`. Should be fixed in GNU binutils > +# 2.35.2. https://sourceware.org/bugzilla/show_bug.cgi?id=25611 > +echo '.file 0 "filename" md5 0x7a0b65214090b6693bd1dc24dd248245' | \ > + $* -gdwarf-5 -Wno-unused-command-line-argument -c -x assembler -o /dev/null - > -- > 2.30.0.365.g02bc693789-goog >
On Sun, Jan 31, 2021 at 12:10 AM Sedat Dilek <sedat.dilek@gmail.com> wrote: > > On Sat, Jan 30, 2021 at 1:44 AM Nick Desaulniers > <ndesaulniers@google.com> wrote: > > > > DWARF v5 is the latest standard of the DWARF debug info format. > > > > Feature detection of DWARF5 is onerous, especially given that we've > > removed $(AS), so we must query $(CC) for DWARF5 assembler directive > > support. > > > > The DWARF version of a binary can be validated with: > > $ llvm-dwarfdump vmlinux | head -n 4 | grep version > > or > > $ readelf --debug-dump=info vmlinux 2>/dev/null | grep Version > > > > DWARF5 wins significantly in terms of size when mixed with compression > > (CONFIG_DEBUG_INFO_COMPRESSED). > > > > 363M vmlinux.clang12.dwarf5.compressed > > 434M vmlinux.clang12.dwarf4.compressed > > 439M vmlinux.clang12.dwarf2.compressed > > 457M vmlinux.clang12.dwarf5 > > 536M vmlinux.clang12.dwarf4 > > 548M vmlinux.clang12.dwarf2 > > > > 515M vmlinux.gcc10.2.dwarf5.compressed > > 599M vmlinux.gcc10.2.dwarf4.compressed > > 624M vmlinux.gcc10.2.dwarf2.compressed > > 630M vmlinux.gcc10.2.dwarf5 > > 765M vmlinux.gcc10.2.dwarf4 > > 809M vmlinux.gcc10.2.dwarf2 > > > > Though the quality of debug info is harder to quantify; size is not a > > proxy for quality. > > > > Jakub notes: > > All [GCC] 5.1 - 6.x did was start accepting -gdwarf-5 as experimental > > option that enabled some small DWARF subset (initially only a few > > DW_LANG_* codes newly added to DWARF5 drafts). Only GCC 7 (released > > after DWARF 5 has been finalized) started emitting DWARF5 section > > headers and got most of the DWARF5 changes in... > > > > Version check GCC so that we don't need to worry about the difference in > > command line args between GNU readelf and llvm-readelf/llvm-dwarfdump to > > validate the DWARF Version in the assembler feature detection script. > > > > GNU `as` only recently gained support for specifying -gdwarf-5, so when > > compiling with Clang but without Clang's integrated assembler > > (LLVM_IAS=1 is not set), explicitly add -Wa,-gdwarf-5 to DEBUG_CFLAGS. > > > > Disabled for now if CONFIG_DEBUG_INFO_BTF is set; pahole doesn't yet > > recognize the new additions to the DWARF debug info. Thanks to Sedat for > > the report. > > > > Link: http://www.dwarfstd.org/doc/DWARF5.pdf > > Reported-by: Sedat Dilek <sedat.dilek@gmail.com> > > Suggested-by: Arvind Sankar <nivedita@alum.mit.edu> > > Suggested-by: Caroline Tice <cmtice@google.com> > > Suggested-by: Fangrui Song <maskray@google.com> > > Suggested-by: Jakub Jelinek <jakub@redhat.com> > > Suggested-by: Masahiro Yamada <masahiroy@kernel.org> > > Suggested-by: Nathan Chancellor <natechancellor@gmail.com> > > Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> > > --- > > Makefile | 1 + > > include/asm-generic/vmlinux.lds.h | 7 ++++++- > > lib/Kconfig.debug | 18 ++++++++++++++++++ > > scripts/test_dwarf5_support.sh | 8 ++++++++ > > 4 files changed, 33 insertions(+), 1 deletion(-) > > create mode 100755 scripts/test_dwarf5_support.sh > > > > diff --git a/Makefile b/Makefile > > index d2b4980807e0..5387a6f2f62d 100644 > > --- a/Makefile > > +++ b/Makefile > > @@ -831,6 +831,7 @@ KBUILD_AFLAGS += -Wa,-gdwarf-2 > > endif > > > > dwarf-version-$(CONFIG_DEBUG_INFO_DWARF4) := 4 > > +dwarf-version-$(CONFIG_DEBUG_INFO_DWARF5) := 5 > > DEBUG_CFLAGS += -gdwarf-$(dwarf-version-y) > > > > ifdef CONFIG_DEBUG_INFO_REDUCED > > diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h > > index 34b7e0d2346c..1e7cde4bd3f9 100644 > > --- a/include/asm-generic/vmlinux.lds.h > > +++ b/include/asm-generic/vmlinux.lds.h > > @@ -842,8 +842,13 @@ > > /* DWARF 4 */ \ > > .debug_types 0 : { *(.debug_types) } \ > > /* DWARF 5 */ \ > > + .debug_addr 0 : { *(.debug_addr) } \ > > + .debug_line_str 0 : { *(.debug_line_str) } \ > > + .debug_loclists 0 : { *(.debug_loclists) } \ > > .debug_macro 0 : { *(.debug_macro) } \ > > - .debug_addr 0 : { *(.debug_addr) } > > + .debug_names 0 : { *(.debug_names) } \ > > + .debug_rnglists 0 : { *(.debug_rnglists) } \ > > + .debug_str_offsets 0 : { *(.debug_str_offsets) } > > > > I just looked at binutils 2.36 in the Debian/experimental repositories. > > [1] says: > > + PR ld/27230 > + * scripttempl/DWARF.sc: Add DWARF-5 .debug_* sections. > > ... > > - /* DWARF Extension. */ > - .debug_macro 0 : { *(.debug_macro) } > + /* DWARF 5. */ > .debug_addr 0 : { *(.debug_addr) } > + .debug_line_str 0 : { *(.debug_line_str) } > + .debug_loclists 0 : { *(.debug_loclists) } > + .debug_macro 0 : { *(.debug_macro) } > + .debug_names 0 : { *(.debug_names) } > + .debug_rnglists 0 : { *(.debug_rnglists) } > + .debug_str_offsets 0 : { *(.debug_str_offsets) } > + .debug_sup 0 : { *(.debug_sup) } > > The list of DWARF-5 .debug_* sections is alphabetically sorted. > AFAICS .debug_sup section is missing? > Here the link to PR ld/27230 ("[PATCH] Add DWARF v5 sections and v4 .debug_types"). - Sedat - [1] https://sourceware.org/pipermail/binutils/2021-January/115064.html > - Sedat - > > [1] https://salsa.debian.org/toolchain-team/binutils/-/commit/f58f33081035672b01a04326a9c8daadbd09a430 > > > /* Stabs debugging sections. */ > > #define STABS_DEBUG \ > > diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug > > index 94c1a7ed6306..ad6f78989d4f 100644 > > --- a/lib/Kconfig.debug > > +++ b/lib/Kconfig.debug > > @@ -268,6 +268,24 @@ config DEBUG_INFO_DWARF4 > > It makes the debug information larger, but it significantly > > improves the success of resolving variables in gdb on optimized code. > > > > +config DEBUG_INFO_DWARF5 > > + bool "Generate DWARF Version 5 debuginfo" > > + depends on GCC_VERSION >= 50000 || CC_IS_CLANG > > + depends on CC_IS_GCC || $(success,$(srctree)/scripts/test_dwarf5_support.sh $(CC) $(CLANG_FLAGS)) > > + depends on !DEBUG_INFO_BTF > > + help > > + Generate DWARF v5 debug info. Requires binutils 2.35.2, gcc 5.0+ (gcc > > + 5.0+ accepts the -gdwarf-5 flag but only had partial support for some > > + draft features until 7.0), and gdb 8.0+. > > + > > + Changes to the structure of debug info in Version 5 allow for around > > + 15-18% savings in resulting image and debug info section sizes as > > + compared to DWARF Version 4. DWARF Version 5 standardizes previous > > + extensions such as accelerators for symbol indexing and the format > > + for fission (.dwo/.dwp) files. Users may not want to select this > > + config if they rely on tooling that has not yet been updated to > > + support DWARF Version 5. > > + > > endchoice # "DWARF version" > > > > config DEBUG_INFO_BTF > > diff --git a/scripts/test_dwarf5_support.sh b/scripts/test_dwarf5_support.sh > > new file mode 100755 > > index 000000000000..c46e2456b47a > > --- /dev/null > > +++ b/scripts/test_dwarf5_support.sh > > @@ -0,0 +1,8 @@ > > +#!/bin/sh > > +# SPDX-License-Identifier: GPL-2.0 > > + > > +# Test that the assembler doesn't need -Wa,-gdwarf-5 when presented with DWARF > > +# v5 input, such as `.file 0` and `md5 0x00`. Should be fixed in GNU binutils > > +# 2.35.2. https://sourceware.org/bugzilla/show_bug.cgi?id=25611 > > +echo '.file 0 "filename" md5 0x7a0b65214090b6693bd1dc24dd248245' | \ > > + $* -gdwarf-5 -Wno-unused-command-line-argument -c -x assembler -o /dev/null - > > -- > > 2.30.0.365.g02bc693789-goog > >
On Sat, Jan 30, 2021 at 3:10 PM Sedat Dilek <sedat.dilek@gmail.com> wrote: > > On Sat, Jan 30, 2021 at 1:44 AM Nick Desaulniers > <ndesaulniers@google.com> wrote: > > > > DWARF v5 is the latest standard of the DWARF debug info format. > > > > Feature detection of DWARF5 is onerous, especially given that we've > > removed $(AS), so we must query $(CC) for DWARF5 assembler directive > > support. > > > > The DWARF version of a binary can be validated with: > > $ llvm-dwarfdump vmlinux | head -n 4 | grep version > > or > > $ readelf --debug-dump=info vmlinux 2>/dev/null | grep Version > > > > DWARF5 wins significantly in terms of size when mixed with compression > > (CONFIG_DEBUG_INFO_COMPRESSED). > > > > 363M vmlinux.clang12.dwarf5.compressed > > 434M vmlinux.clang12.dwarf4.compressed > > 439M vmlinux.clang12.dwarf2.compressed > > 457M vmlinux.clang12.dwarf5 > > 536M vmlinux.clang12.dwarf4 > > 548M vmlinux.clang12.dwarf2 > > > > 515M vmlinux.gcc10.2.dwarf5.compressed > > 599M vmlinux.gcc10.2.dwarf4.compressed > > 624M vmlinux.gcc10.2.dwarf2.compressed > > 630M vmlinux.gcc10.2.dwarf5 > > 765M vmlinux.gcc10.2.dwarf4 > > 809M vmlinux.gcc10.2.dwarf2 > > > > Though the quality of debug info is harder to quantify; size is not a > > proxy for quality. > > > > Jakub notes: > > All [GCC] 5.1 - 6.x did was start accepting -gdwarf-5 as experimental > > option that enabled some small DWARF subset (initially only a few > > DW_LANG_* codes newly added to DWARF5 drafts). Only GCC 7 (released > > after DWARF 5 has been finalized) started emitting DWARF5 section > > headers and got most of the DWARF5 changes in... > > > > Version check GCC so that we don't need to worry about the difference in > > command line args between GNU readelf and llvm-readelf/llvm-dwarfdump to > > validate the DWARF Version in the assembler feature detection script. > > > > GNU `as` only recently gained support for specifying -gdwarf-5, so when > > compiling with Clang but without Clang's integrated assembler > > (LLVM_IAS=1 is not set), explicitly add -Wa,-gdwarf-5 to DEBUG_CFLAGS. > > > > Disabled for now if CONFIG_DEBUG_INFO_BTF is set; pahole doesn't yet > > recognize the new additions to the DWARF debug info. Thanks to Sedat for > > the report. > > > > Link: http://www.dwarfstd.org/doc/DWARF5.pdf > > Reported-by: Sedat Dilek <sedat.dilek@gmail.com> > > Suggested-by: Arvind Sankar <nivedita@alum.mit.edu> > > Suggested-by: Caroline Tice <cmtice@google.com> > > Suggested-by: Fangrui Song <maskray@google.com> > > Suggested-by: Jakub Jelinek <jakub@redhat.com> > > Suggested-by: Masahiro Yamada <masahiroy@kernel.org> > > Suggested-by: Nathan Chancellor <natechancellor@gmail.com> > > Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> > > --- > > Makefile | 1 + > > include/asm-generic/vmlinux.lds.h | 7 ++++++- > > lib/Kconfig.debug | 18 ++++++++++++++++++ > > scripts/test_dwarf5_support.sh | 8 ++++++++ > > 4 files changed, 33 insertions(+), 1 deletion(-) > > create mode 100755 scripts/test_dwarf5_support.sh > > > > diff --git a/Makefile b/Makefile > > index d2b4980807e0..5387a6f2f62d 100644 > > --- a/Makefile > > +++ b/Makefile > > @@ -831,6 +831,7 @@ KBUILD_AFLAGS += -Wa,-gdwarf-2 > > endif > > > > dwarf-version-$(CONFIG_DEBUG_INFO_DWARF4) := 4 > > +dwarf-version-$(CONFIG_DEBUG_INFO_DWARF5) := 5 > > DEBUG_CFLAGS += -gdwarf-$(dwarf-version-y) > > > > ifdef CONFIG_DEBUG_INFO_REDUCED > > diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h > > index 34b7e0d2346c..1e7cde4bd3f9 100644 > > --- a/include/asm-generic/vmlinux.lds.h > > +++ b/include/asm-generic/vmlinux.lds.h > > @@ -842,8 +842,13 @@ > > /* DWARF 4 */ \ > > .debug_types 0 : { *(.debug_types) } \ > > /* DWARF 5 */ \ > > + .debug_addr 0 : { *(.debug_addr) } \ > > + .debug_line_str 0 : { *(.debug_line_str) } \ > > + .debug_loclists 0 : { *(.debug_loclists) } \ > > .debug_macro 0 : { *(.debug_macro) } \ > > - .debug_addr 0 : { *(.debug_addr) } > > + .debug_names 0 : { *(.debug_names) } \ > > + .debug_rnglists 0 : { *(.debug_rnglists) } \ > > + .debug_str_offsets 0 : { *(.debug_str_offsets) } > > > > I just looked at binutils 2.36 in the Debian/experimental repositories. > > [1] says: > > + PR ld/27230 > + * scripttempl/DWARF.sc: Add DWARF-5 .debug_* sections. > > ... > > - /* DWARF Extension. */ > - .debug_macro 0 : { *(.debug_macro) } > + /* DWARF 5. */ > .debug_addr 0 : { *(.debug_addr) } > + .debug_line_str 0 : { *(.debug_line_str) } > + .debug_loclists 0 : { *(.debug_loclists) } > + .debug_macro 0 : { *(.debug_macro) } > + .debug_names 0 : { *(.debug_names) } > + .debug_rnglists 0 : { *(.debug_rnglists) } > + .debug_str_offsets 0 : { *(.debug_str_offsets) } > + .debug_sup 0 : { *(.debug_sup) } > > The list of DWARF-5 .debug_* sections is alphabetically sorted. > AFAICS .debug_sup section is missing? > > - Sedat - No compiler produces .debug_sup section. It could be from some post-processing tool which is unrelated to the linker. Omitting it is fine. > [1] https://salsa.debian.org/toolchain-team/binutils/-/commit/f58f33081035672b01a04326a9c8daadbd09a430 > > > /* Stabs debugging sections. */ > > #define STABS_DEBUG \ > > diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug > > index 94c1a7ed6306..ad6f78989d4f 100644 > > --- a/lib/Kconfig.debug > > +++ b/lib/Kconfig.debug > > @@ -268,6 +268,24 @@ config DEBUG_INFO_DWARF4 > > It makes the debug information larger, but it significantly > > improves the success of resolving variables in gdb on optimized code. > > > > +config DEBUG_INFO_DWARF5 > > + bool "Generate DWARF Version 5 debuginfo" > > + depends on GCC_VERSION >= 50000 || CC_IS_CLANG > > + depends on CC_IS_GCC || $(success,$(srctree)/scripts/test_dwarf5_support.sh $(CC) $(CLANG_FLAGS)) > > + depends on !DEBUG_INFO_BTF > > + help > > + Generate DWARF v5 debug info. Requires binutils 2.35.2, gcc 5.0+ (gcc > > + 5.0+ accepts the -gdwarf-5 flag but only had partial support for some > > + draft features until 7.0), and gdb 8.0+. > > + > > + Changes to the structure of debug info in Version 5 allow for around > > + 15-18% savings in resulting image and debug info section sizes as > > + compared to DWARF Version 4. DWARF Version 5 standardizes previous > > + extensions such as accelerators for symbol indexing and the format > > + for fission (.dwo/.dwp) files. Users may not want to select this > > + config if they rely on tooling that has not yet been updated to > > + support DWARF Version 5. > > + > > endchoice # "DWARF version" > > > > config DEBUG_INFO_BTF > > diff --git a/scripts/test_dwarf5_support.sh b/scripts/test_dwarf5_support.sh > > new file mode 100755 > > index 000000000000..c46e2456b47a > > --- /dev/null > > +++ b/scripts/test_dwarf5_support.sh > > @@ -0,0 +1,8 @@ > > +#!/bin/sh > > +# SPDX-License-Identifier: GPL-2.0 > > + > > +# Test that the assembler doesn't need -Wa,-gdwarf-5 when presented with DWARF > > +# v5 input, such as `.file 0` and `md5 0x00`. Should be fixed in GNU binutils > > +# 2.35.2. https://sourceware.org/bugzilla/show_bug.cgi?id=25611 > > +echo '.file 0 "filename" md5 0x7a0b65214090b6693bd1dc24dd248245' | \ > > + $* -gdwarf-5 -Wno-unused-command-line-argument -c -x assembler -o /dev/null - > > -- > > 2.30.0.365.g02bc693789-goog > >
On Sun, Jan 31, 2021 at 1:37 AM Fāng-ruì Sòng <maskray@google.com> wrote: > > On Sat, Jan 30, 2021 at 3:10 PM Sedat Dilek <sedat.dilek@gmail.com> wrote: > > > > On Sat, Jan 30, 2021 at 1:44 AM Nick Desaulniers > > <ndesaulniers@google.com> wrote: > > > > > > DWARF v5 is the latest standard of the DWARF debug info format. > > > > > > Feature detection of DWARF5 is onerous, especially given that we've > > > removed $(AS), so we must query $(CC) for DWARF5 assembler directive > > > support. > > > > > > The DWARF version of a binary can be validated with: > > > $ llvm-dwarfdump vmlinux | head -n 4 | grep version > > > or > > > $ readelf --debug-dump=info vmlinux 2>/dev/null | grep Version > > > > > > DWARF5 wins significantly in terms of size when mixed with compression > > > (CONFIG_DEBUG_INFO_COMPRESSED). > > > > > > 363M vmlinux.clang12.dwarf5.compressed > > > 434M vmlinux.clang12.dwarf4.compressed > > > 439M vmlinux.clang12.dwarf2.compressed > > > 457M vmlinux.clang12.dwarf5 > > > 536M vmlinux.clang12.dwarf4 > > > 548M vmlinux.clang12.dwarf2 > > > > > > 515M vmlinux.gcc10.2.dwarf5.compressed > > > 599M vmlinux.gcc10.2.dwarf4.compressed > > > 624M vmlinux.gcc10.2.dwarf2.compressed > > > 630M vmlinux.gcc10.2.dwarf5 > > > 765M vmlinux.gcc10.2.dwarf4 > > > 809M vmlinux.gcc10.2.dwarf2 > > > > > > Though the quality of debug info is harder to quantify; size is not a > > > proxy for quality. > > > > > > Jakub notes: > > > All [GCC] 5.1 - 6.x did was start accepting -gdwarf-5 as experimental > > > option that enabled some small DWARF subset (initially only a few > > > DW_LANG_* codes newly added to DWARF5 drafts). Only GCC 7 (released > > > after DWARF 5 has been finalized) started emitting DWARF5 section > > > headers and got most of the DWARF5 changes in... > > > > > > Version check GCC so that we don't need to worry about the difference in > > > command line args between GNU readelf and llvm-readelf/llvm-dwarfdump to > > > validate the DWARF Version in the assembler feature detection script. > > > > > > GNU `as` only recently gained support for specifying -gdwarf-5, so when > > > compiling with Clang but without Clang's integrated assembler > > > (LLVM_IAS=1 is not set), explicitly add -Wa,-gdwarf-5 to DEBUG_CFLAGS. > > > > > > Disabled for now if CONFIG_DEBUG_INFO_BTF is set; pahole doesn't yet > > > recognize the new additions to the DWARF debug info. Thanks to Sedat for > > > the report. > > > > > > Link: http://www.dwarfstd.org/doc/DWARF5.pdf > > > Reported-by: Sedat Dilek <sedat.dilek@gmail.com> > > > Suggested-by: Arvind Sankar <nivedita@alum.mit.edu> > > > Suggested-by: Caroline Tice <cmtice@google.com> > > > Suggested-by: Fangrui Song <maskray@google.com> > > > Suggested-by: Jakub Jelinek <jakub@redhat.com> > > > Suggested-by: Masahiro Yamada <masahiroy@kernel.org> > > > Suggested-by: Nathan Chancellor <natechancellor@gmail.com> > > > Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> > > > --- > > > Makefile | 1 + > > > include/asm-generic/vmlinux.lds.h | 7 ++++++- > > > lib/Kconfig.debug | 18 ++++++++++++++++++ > > > scripts/test_dwarf5_support.sh | 8 ++++++++ > > > 4 files changed, 33 insertions(+), 1 deletion(-) > > > create mode 100755 scripts/test_dwarf5_support.sh > > > > > > diff --git a/Makefile b/Makefile > > > index d2b4980807e0..5387a6f2f62d 100644 > > > --- a/Makefile > > > +++ b/Makefile > > > @@ -831,6 +831,7 @@ KBUILD_AFLAGS += -Wa,-gdwarf-2 > > > endif > > > > > > dwarf-version-$(CONFIG_DEBUG_INFO_DWARF4) := 4 > > > +dwarf-version-$(CONFIG_DEBUG_INFO_DWARF5) := 5 > > > DEBUG_CFLAGS += -gdwarf-$(dwarf-version-y) > > > > > > ifdef CONFIG_DEBUG_INFO_REDUCED > > > diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h > > > index 34b7e0d2346c..1e7cde4bd3f9 100644 > > > --- a/include/asm-generic/vmlinux.lds.h > > > +++ b/include/asm-generic/vmlinux.lds.h > > > @@ -842,8 +842,13 @@ > > > /* DWARF 4 */ \ > > > .debug_types 0 : { *(.debug_types) } \ > > > /* DWARF 5 */ \ > > > + .debug_addr 0 : { *(.debug_addr) } \ > > > + .debug_line_str 0 : { *(.debug_line_str) } \ > > > + .debug_loclists 0 : { *(.debug_loclists) } \ > > > .debug_macro 0 : { *(.debug_macro) } \ > > > - .debug_addr 0 : { *(.debug_addr) } > > > + .debug_names 0 : { *(.debug_names) } \ > > > + .debug_rnglists 0 : { *(.debug_rnglists) } \ > > > + .debug_str_offsets 0 : { *(.debug_str_offsets) } > > > > > > > I just looked at binutils 2.36 in the Debian/experimental repositories. > > > > [1] says: > > > > + PR ld/27230 > > + * scripttempl/DWARF.sc: Add DWARF-5 .debug_* sections. > > > > ... > > > > - /* DWARF Extension. */ > > - .debug_macro 0 : { *(.debug_macro) } > > + /* DWARF 5. */ > > .debug_addr 0 : { *(.debug_addr) } > > + .debug_line_str 0 : { *(.debug_line_str) } > > + .debug_loclists 0 : { *(.debug_loclists) } > > + .debug_macro 0 : { *(.debug_macro) } > > + .debug_names 0 : { *(.debug_names) } > > + .debug_rnglists 0 : { *(.debug_rnglists) } > > + .debug_str_offsets 0 : { *(.debug_str_offsets) } > > + .debug_sup 0 : { *(.debug_sup) } > > > > The list of DWARF-5 .debug_* sections is alphabetically sorted. > > AFAICS .debug_sup section is missing? > > > > - Sedat - > > No compiler produces .debug_sup section. It could be from some > post-processing tool which is unrelated to the linker. > Omitting it is fine. > Thanks for the clarification, Fangrui. - Sedat - > > [1] https://salsa.debian.org/toolchain-team/binutils/-/commit/f58f33081035672b01a04326a9c8daadbd09a430 > > > > > /* Stabs debugging sections. */ > > > #define STABS_DEBUG \ > > > diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug > > > index 94c1a7ed6306..ad6f78989d4f 100644 > > > --- a/lib/Kconfig.debug > > > +++ b/lib/Kconfig.debug > > > @@ -268,6 +268,24 @@ config DEBUG_INFO_DWARF4 > > > It makes the debug information larger, but it significantly > > > improves the success of resolving variables in gdb on optimized code. > > > > > > +config DEBUG_INFO_DWARF5 > > > + bool "Generate DWARF Version 5 debuginfo" > > > + depends on GCC_VERSION >= 50000 || CC_IS_CLANG > > > + depends on CC_IS_GCC || $(success,$(srctree)/scripts/test_dwarf5_support.sh $(CC) $(CLANG_FLAGS)) > > > + depends on !DEBUG_INFO_BTF > > > + help > > > + Generate DWARF v5 debug info. Requires binutils 2.35.2, gcc 5.0+ (gcc > > > + 5.0+ accepts the -gdwarf-5 flag but only had partial support for some > > > + draft features until 7.0), and gdb 8.0+. > > > + > > > + Changes to the structure of debug info in Version 5 allow for around > > > + 15-18% savings in resulting image and debug info section sizes as > > > + compared to DWARF Version 4. DWARF Version 5 standardizes previous > > > + extensions such as accelerators for symbol indexing and the format > > > + for fission (.dwo/.dwp) files. Users may not want to select this > > > + config if they rely on tooling that has not yet been updated to > > > + support DWARF Version 5. > > > + > > > endchoice # "DWARF version" > > > > > > config DEBUG_INFO_BTF > > > diff --git a/scripts/test_dwarf5_support.sh b/scripts/test_dwarf5_support.sh > > > new file mode 100755 > > > index 000000000000..c46e2456b47a > > > --- /dev/null > > > +++ b/scripts/test_dwarf5_support.sh > > > @@ -0,0 +1,8 @@ > > > +#!/bin/sh > > > +# SPDX-License-Identifier: GPL-2.0 > > > + > > > +# Test that the assembler doesn't need -Wa,-gdwarf-5 when presented with DWARF > > > +# v5 input, such as `.file 0` and `md5 0x00`. Should be fixed in GNU binutils > > > +# 2.35.2. https://sourceware.org/bugzilla/show_bug.cgi?id=25611 > > > +echo '.file 0 "filename" md5 0x7a0b65214090b6693bd1dc24dd248245' | \ > > > + $* -gdwarf-5 -Wno-unused-command-line-argument -c -x assembler -o /dev/null - > > > -- > > > 2.30.0.365.g02bc693789-goog > > >
On Sat, Jan 30, 2021 at 9:44 AM 'Nick Desaulniers' via Clang Built Linux <clang-built-linux@googlegroups.com> wrote: > > DWARF v5 is the latest standard of the DWARF debug info format. > > Feature detection of DWARF5 is onerous, especially given that we've > removed $(AS), so we must query $(CC) for DWARF5 assembler directive > support. > > The DWARF version of a binary can be validated with: > $ llvm-dwarfdump vmlinux | head -n 4 | grep version > or > $ readelf --debug-dump=info vmlinux 2>/dev/null | grep Version > > DWARF5 wins significantly in terms of size when mixed with compression > (CONFIG_DEBUG_INFO_COMPRESSED). > > 363M vmlinux.clang12.dwarf5.compressed > 434M vmlinux.clang12.dwarf4.compressed > 439M vmlinux.clang12.dwarf2.compressed > 457M vmlinux.clang12.dwarf5 > 536M vmlinux.clang12.dwarf4 > 548M vmlinux.clang12.dwarf2 > > 515M vmlinux.gcc10.2.dwarf5.compressed > 599M vmlinux.gcc10.2.dwarf4.compressed > 624M vmlinux.gcc10.2.dwarf2.compressed > 630M vmlinux.gcc10.2.dwarf5 > 765M vmlinux.gcc10.2.dwarf4 > 809M vmlinux.gcc10.2.dwarf2 > > Though the quality of debug info is harder to quantify; size is not a > proxy for quality. > > Jakub notes: > All [GCC] 5.1 - 6.x did was start accepting -gdwarf-5 as experimental > option that enabled some small DWARF subset (initially only a few > DW_LANG_* codes newly added to DWARF5 drafts). Only GCC 7 (released > after DWARF 5 has been finalized) started emitting DWARF5 section > headers and got most of the DWARF5 changes in... > > Version check GCC so that we don't need to worry about the difference in > command line args between GNU readelf and llvm-readelf/llvm-dwarfdump to > validate the DWARF Version in the assembler feature detection script. > > GNU `as` only recently gained support for specifying -gdwarf-5, so when > compiling with Clang but without Clang's integrated assembler > (LLVM_IAS=1 is not set), explicitly add -Wa,-gdwarf-5 to DEBUG_CFLAGS. > > Disabled for now if CONFIG_DEBUG_INFO_BTF is set; pahole doesn't yet > recognize the new additions to the DWARF debug info. Thanks to Sedat for > the report. > > Link: http://www.dwarfstd.org/doc/DWARF5.pdf > Reported-by: Sedat Dilek <sedat.dilek@gmail.com> > Suggested-by: Arvind Sankar <nivedita@alum.mit.edu> > Suggested-by: Caroline Tice <cmtice@google.com> > Suggested-by: Fangrui Song <maskray@google.com> > Suggested-by: Jakub Jelinek <jakub@redhat.com> > Suggested-by: Masahiro Yamada <masahiroy@kernel.org> > Suggested-by: Nathan Chancellor <natechancellor@gmail.com> > Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> > --- > Makefile | 1 + > include/asm-generic/vmlinux.lds.h | 7 ++++++- > lib/Kconfig.debug | 18 ++++++++++++++++++ > scripts/test_dwarf5_support.sh | 8 ++++++++ > 4 files changed, 33 insertions(+), 1 deletion(-) > create mode 100755 scripts/test_dwarf5_support.sh Nick, the patch set is getting simpler and simpler, and almost good enough to be merged. Please let me ask two questions below. There has been a lot of discussion, and I might have missed the context. > diff --git a/Makefile b/Makefile > index d2b4980807e0..5387a6f2f62d 100644 > --- a/Makefile > +++ b/Makefile > @@ -831,6 +831,7 @@ KBUILD_AFLAGS += -Wa,-gdwarf-2 > endif > > dwarf-version-$(CONFIG_DEBUG_INFO_DWARF4) := 4 > +dwarf-version-$(CONFIG_DEBUG_INFO_DWARF5) := 5 > DEBUG_CFLAGS += -gdwarf-$(dwarf-version-y) > > ifdef CONFIG_DEBUG_INFO_REDUCED > diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h > index 34b7e0d2346c..1e7cde4bd3f9 100644 > --- a/include/asm-generic/vmlinux.lds.h > +++ b/include/asm-generic/vmlinux.lds.h > @@ -842,8 +842,13 @@ > /* DWARF 4 */ \ > .debug_types 0 : { *(.debug_types) } \ > /* DWARF 5 */ \ > + .debug_addr 0 : { *(.debug_addr) } \ > + .debug_line_str 0 : { *(.debug_line_str) } \ > + .debug_loclists 0 : { *(.debug_loclists) } \ > .debug_macro 0 : { *(.debug_macro) } \ > - .debug_addr 0 : { *(.debug_addr) } > + .debug_names 0 : { *(.debug_names) } \ > + .debug_rnglists 0 : { *(.debug_rnglists) } \ > + .debug_str_offsets 0 : { *(.debug_str_offsets) } > > /* Stabs debugging sections. */ > #define STABS_DEBUG \ > diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug > index 94c1a7ed6306..ad6f78989d4f 100644 > --- a/lib/Kconfig.debug > +++ b/lib/Kconfig.debug > @@ -268,6 +268,24 @@ config DEBUG_INFO_DWARF4 > It makes the debug information larger, but it significantly > improves the success of resolving variables in gdb on optimized code. > > +config DEBUG_INFO_DWARF5 > + bool "Generate DWARF Version 5 debuginfo" > + depends on GCC_VERSION >= 50000 || CC_IS_CLANG > + depends on CC_IS_GCC || $(success,$(srctree)/scripts/test_dwarf5_support.sh $(CC) $(CLANG_FLAGS)) Q1. This "CC_IS_GCC ||" was introduced by v4. GCC never outputs '.file 0', which is why this test is only needed for Clang, correct? > + depends on !DEBUG_INFO_BTF > + help > + Generate DWARF v5 debug info. Requires binutils 2.35.2, gcc 5.0+ (gcc > + 5.0+ accepts the -gdwarf-5 flag but only had partial support for some > + draft features until 7.0), and gdb 8.0+. > + > + Changes to the structure of debug info in Version 5 allow for around > + 15-18% savings in resulting image and debug info section sizes as > + compared to DWARF Version 4. DWARF Version 5 standardizes previous > + extensions such as accelerators for symbol indexing and the format > + for fission (.dwo/.dwp) files. Users may not want to select this > + config if they rely on tooling that has not yet been updated to > + support DWARF Version 5. > + > endchoice # "DWARF version" > > config DEBUG_INFO_BTF > diff --git a/scripts/test_dwarf5_support.sh b/scripts/test_dwarf5_support.sh > new file mode 100755 > index 000000000000..c46e2456b47a > --- /dev/null > +++ b/scripts/test_dwarf5_support.sh > @@ -0,0 +1,8 @@ > +#!/bin/sh > +# SPDX-License-Identifier: GPL-2.0 > + > +# Test that the assembler doesn't need -Wa,-gdwarf-5 when presented with DWARF > +# v5 input, such as `.file 0` and `md5 0x00`. Should be fixed in GNU binutils > +# 2.35.2. https://sourceware.org/bugzilla/show_bug.cgi?id=25611 I saw the following links in v6. https://sourceware.org/bugzilla/show_bug.cgi?id=25612 https://sourceware.org/bugzilla/show_bug.cgi?id=25614 They were dropped in v7. Why? I just thought they were good to know... > +echo '.file 0 "filename" md5 0x7a0b65214090b6693bd1dc24dd248245' | \ > + $* -gdwarf-5 -Wno-unused-command-line-argument -c -x assembler -o /dev/null - > -- > 2.30.0.365.g02bc693789-goog > > -- > You received this message because you are subscribed to the Google Groups "Clang Built Linux" group. > To unsubscribe from this group and stop receiving emails from it, send an email to clang-built-linux+unsubscribe@googlegroups.com. > To view this discussion on the web visit https://groups.google.com/d/msgid/clang-built-linux/20210130004401.2528717-3-ndesaulniers%40google.com.
On Wed, Feb 3, 2021 at 3:07 PM Masahiro Yamada <masahiroy@kernel.org> wrote: > > Nick, the patch set is getting simpler and simpler, > and almost good enough to be merged. I agree. I think Sedat pointed out a binutils 2.35.2 release; thanks to Nick Clifton for that. > > > Please let me ask two questions below. > > There has been a lot of discussion, and > I might have missed the context. > > > --- a/lib/Kconfig.debug > > +++ b/lib/Kconfig.debug > > @@ -268,6 +268,24 @@ config DEBUG_INFO_DWARF4 > > It makes the debug information larger, but it significantly > > improves the success of resolving variables in gdb on optimized code. > > > > +config DEBUG_INFO_DWARF5 > > + bool "Generate DWARF Version 5 debuginfo" > > + depends on GCC_VERSION >= 50000 || CC_IS_CLANG > > + depends on CC_IS_GCC || $(success,$(srctree)/scripts/test_dwarf5_support.sh $(CC) $(CLANG_FLAGS)) > > Q1. > > This "CC_IS_GCC ||" was introduced by v4. > > GCC never outputs '.file 0', which is why > this test is only needed for Clang, correct? This test script is only needed when compiling with clang but without its integrated assembler. It checks that when clang is used as the driver, but GAS is used as the assembler, that GAS will be able to decode the DWARF v5 assembler additions Clang will produce without needing an explicit -Wa,-gdwarf-5 flag passed. Technically, it is unnecessary for `LLVM=1 LLVM_IAS=1` or `CC=clang LLVM_IAS=1` (ie. clang+clang's integrated assembler). But there is no way to express AS_IS_IAS today in KConfig (similar to CC_IS_{GCC|CLANG} or LD_IS_LLD). I don't think that's necessary; whether or not clang's integrated assembler is used, when using clang, run the simple check. > > --- /dev/null > > +++ b/scripts/test_dwarf5_support.sh > > @@ -0,0 +1,8 @@ > > +#!/bin/sh > > +# SPDX-License-Identifier: GPL-2.0 > > + > > +# Test that the assembler doesn't need -Wa,-gdwarf-5 when presented with DWARF > > +# v5 input, such as `.file 0` and `md5 0x00`. Should be fixed in GNU binutils > > +# 2.35.2. https://sourceware.org/bugzilla/show_bug.cgi?id=25611 > > > I saw the following links in v6. > > https://sourceware.org/bugzilla/show_bug.cgi?id=25612 > https://sourceware.org/bugzilla/show_bug.cgi?id=25614 > > They were dropped in v7. Why? > > I just thought they were good to know... While having fixes for those bugs is required, technically https://sourceware.org/bugzilla/show_bug.cgi?id=25611 is the latest bug which was fixed. Testing for a fix of https://sourceware.org/bugzilla/show_bug.cgi?id=25611 implies that fixes for 25612 and 25614 exist due to the order they were fixed in GAS. Technically, you could argue that this script is quite GAS centric; given an arbitrary "assembler" the test should check a few things. Realistically, I think that's overkill based on what assemblers are in use today; we can always grow the script should we identify other tests additional assemblers may need to pass, but until then, I suspect YAGNI. Maybe there's a more precise name for the script to reflect that, but that gets close to "what color shall we paint the bikeshed?" Given the number of folks on the thread, plz no.
On Thu, Feb 04, 2021 at 08:06:12AM +0900, Masahiro Yamada wrote: > GCC never outputs '.file 0', which is why > this test is only needed for Clang, correct? No, GCC outputs .file 0 if it during configure time detected assembler that supports it and doesn't have any of the known bugs related to it. But that means kernel doesn't need to care because GCC already took care of that. Jakub
On Thu, Feb 4, 2021 at 8:27 AM Nick Desaulniers <ndesaulniers@google.com> wrote: > > On Wed, Feb 3, 2021 at 3:07 PM Masahiro Yamada <masahiroy@kernel.org> wrote: > > > > Nick, the patch set is getting simpler and simpler, > > and almost good enough to be merged. > > I agree. I think Sedat pointed out a binutils 2.35.2 release; thanks > to Nick Clifton for that. > > > > > > > Please let me ask two questions below. > > > > There has been a lot of discussion, and > > I might have missed the context. > > > > > --- a/lib/Kconfig.debug > > > +++ b/lib/Kconfig.debug > > > @@ -268,6 +268,24 @@ config DEBUG_INFO_DWARF4 > > > It makes the debug information larger, but it significantly > > > improves the success of resolving variables in gdb on optimized code. > > > > > > +config DEBUG_INFO_DWARF5 > > > + bool "Generate DWARF Version 5 debuginfo" > > > + depends on GCC_VERSION >= 50000 || CC_IS_CLANG > > > + depends on CC_IS_GCC || $(success,$(srctree)/scripts/test_dwarf5_support.sh $(CC) $(CLANG_FLAGS)) > > > > Q1. > > > > This "CC_IS_GCC ||" was introduced by v4. > > > > GCC never outputs '.file 0', which is why > > this test is only needed for Clang, correct? > > This test script is only needed when compiling with clang but without > its integrated assembler. It checks that when clang is used as the > driver, but GAS is used as the assembler, that GAS will be able to > decode the DWARF v5 assembler additions Clang will produce without > needing an explicit -Wa,-gdwarf-5 flag passed. > > Technically, it is unnecessary for `LLVM=1 LLVM_IAS=1` or `CC=clang > LLVM_IAS=1` (ie. clang+clang's integrated assembler). But there is no > way to express AS_IS_IAS today in KConfig (similar to > CC_IS_{GCC|CLANG} or LD_IS_LLD). I don't think that's necessary; > whether or not clang's integrated assembler is used, when using clang, > run the simple check. > > > > --- /dev/null > > > +++ b/scripts/test_dwarf5_support.sh > > > @@ -0,0 +1,8 @@ > > > +#!/bin/sh > > > +# SPDX-License-Identifier: GPL-2.0 > > > + > > > +# Test that the assembler doesn't need -Wa,-gdwarf-5 when presented with DWARF > > > +# v5 input, such as `.file 0` and `md5 0x00`. Should be fixed in GNU binutils > > > +# 2.35.2. https://sourceware.org/bugzilla/show_bug.cgi?id=25611 > > > > > > I saw the following links in v6. > > > > https://sourceware.org/bugzilla/show_bug.cgi?id=25612 > > https://sourceware.org/bugzilla/show_bug.cgi?id=25614 > > > > They were dropped in v7. Why? > > > > I just thought they were good to know... > > While having fixes for those bugs is required, technically > https://sourceware.org/bugzilla/show_bug.cgi?id=25611 is the latest > bug which was fixed. Testing for a fix of > https://sourceware.org/bugzilla/show_bug.cgi?id=25611 implies that > fixes for 25612 and 25614 exist due to the order they were fixed in > GAS. It is difficult to know the patch order in the binutils project. Personally, I prefer having all the three references here. Otherwise, it is difficult to understand why this script is doing such complex checks. > Technically, you could argue that this script is quite GAS > centric; given an arbitrary "assembler" the test should check a few > things. Realistically, I think that's overkill based on what > assemblers are in use today; we can always grow the script should we > identify other tests additional assemblers may need to pass, but until > then, I suspect YAGNI. Maybe there's a more precise name for the > script to reflect that, but that gets close to "what color shall we > paint the bikeshed?" Given the number of folks on the thread, plz no. No argument with this regard. I agree with you. > -- > Thanks, > ~Nick Desaulniers
On Thu, Feb 4, 2021 at 8:36 AM Jakub Jelinek <jakub@redhat.com> wrote: > > On Thu, Feb 04, 2021 at 08:06:12AM +0900, Masahiro Yamada wrote: > > GCC never outputs '.file 0', which is why > > this test is only needed for Clang, correct? > > No, GCC outputs .file 0 if it during configure time detected assembler that > supports it and doesn't have any of the known bugs related to it. > But that means kernel doesn't need to care because GCC already took care of > that. OK, then I am fine with the current code. Thanks for the clarification. -- Best Regards Masahiro Yamada
diff --git a/Makefile b/Makefile index d2b4980807e0..5387a6f2f62d 100644 --- a/Makefile +++ b/Makefile @@ -831,6 +831,7 @@ KBUILD_AFLAGS += -Wa,-gdwarf-2 endif dwarf-version-$(CONFIG_DEBUG_INFO_DWARF4) := 4 +dwarf-version-$(CONFIG_DEBUG_INFO_DWARF5) := 5 DEBUG_CFLAGS += -gdwarf-$(dwarf-version-y) ifdef CONFIG_DEBUG_INFO_REDUCED diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h index 34b7e0d2346c..1e7cde4bd3f9 100644 --- a/include/asm-generic/vmlinux.lds.h +++ b/include/asm-generic/vmlinux.lds.h @@ -842,8 +842,13 @@ /* DWARF 4 */ \ .debug_types 0 : { *(.debug_types) } \ /* DWARF 5 */ \ + .debug_addr 0 : { *(.debug_addr) } \ + .debug_line_str 0 : { *(.debug_line_str) } \ + .debug_loclists 0 : { *(.debug_loclists) } \ .debug_macro 0 : { *(.debug_macro) } \ - .debug_addr 0 : { *(.debug_addr) } + .debug_names 0 : { *(.debug_names) } \ + .debug_rnglists 0 : { *(.debug_rnglists) } \ + .debug_str_offsets 0 : { *(.debug_str_offsets) } /* Stabs debugging sections. */ #define STABS_DEBUG \ diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug index 94c1a7ed6306..ad6f78989d4f 100644 --- a/lib/Kconfig.debug +++ b/lib/Kconfig.debug @@ -268,6 +268,24 @@ config DEBUG_INFO_DWARF4 It makes the debug information larger, but it significantly improves the success of resolving variables in gdb on optimized code. +config DEBUG_INFO_DWARF5 + bool "Generate DWARF Version 5 debuginfo" + depends on GCC_VERSION >= 50000 || CC_IS_CLANG + depends on CC_IS_GCC || $(success,$(srctree)/scripts/test_dwarf5_support.sh $(CC) $(CLANG_FLAGS)) + depends on !DEBUG_INFO_BTF + help + Generate DWARF v5 debug info. Requires binutils 2.35.2, gcc 5.0+ (gcc + 5.0+ accepts the -gdwarf-5 flag but only had partial support for some + draft features until 7.0), and gdb 8.0+. + + Changes to the structure of debug info in Version 5 allow for around + 15-18% savings in resulting image and debug info section sizes as + compared to DWARF Version 4. DWARF Version 5 standardizes previous + extensions such as accelerators for symbol indexing and the format + for fission (.dwo/.dwp) files. Users may not want to select this + config if they rely on tooling that has not yet been updated to + support DWARF Version 5. + endchoice # "DWARF version" config DEBUG_INFO_BTF diff --git a/scripts/test_dwarf5_support.sh b/scripts/test_dwarf5_support.sh new file mode 100755 index 000000000000..c46e2456b47a --- /dev/null +++ b/scripts/test_dwarf5_support.sh @@ -0,0 +1,8 @@ +#!/bin/sh +# SPDX-License-Identifier: GPL-2.0 + +# Test that the assembler doesn't need -Wa,-gdwarf-5 when presented with DWARF +# v5 input, such as `.file 0` and `md5 0x00`. Should be fixed in GNU binutils +# 2.35.2. https://sourceware.org/bugzilla/show_bug.cgi?id=25611 +echo '.file 0 "filename" md5 0x7a0b65214090b6693bd1dc24dd248245' | \ + $* -gdwarf-5 -Wno-unused-command-line-argument -c -x assembler -o /dev/null -
DWARF v5 is the latest standard of the DWARF debug info format. Feature detection of DWARF5 is onerous, especially given that we've removed $(AS), so we must query $(CC) for DWARF5 assembler directive support. The DWARF version of a binary can be validated with: $ llvm-dwarfdump vmlinux | head -n 4 | grep version or $ readelf --debug-dump=info vmlinux 2>/dev/null | grep Version DWARF5 wins significantly in terms of size when mixed with compression (CONFIG_DEBUG_INFO_COMPRESSED). 363M vmlinux.clang12.dwarf5.compressed 434M vmlinux.clang12.dwarf4.compressed 439M vmlinux.clang12.dwarf2.compressed 457M vmlinux.clang12.dwarf5 536M vmlinux.clang12.dwarf4 548M vmlinux.clang12.dwarf2 515M vmlinux.gcc10.2.dwarf5.compressed 599M vmlinux.gcc10.2.dwarf4.compressed 624M vmlinux.gcc10.2.dwarf2.compressed 630M vmlinux.gcc10.2.dwarf5 765M vmlinux.gcc10.2.dwarf4 809M vmlinux.gcc10.2.dwarf2 Though the quality of debug info is harder to quantify; size is not a proxy for quality. Jakub notes: All [GCC] 5.1 - 6.x did was start accepting -gdwarf-5 as experimental option that enabled some small DWARF subset (initially only a few DW_LANG_* codes newly added to DWARF5 drafts). Only GCC 7 (released after DWARF 5 has been finalized) started emitting DWARF5 section headers and got most of the DWARF5 changes in... Version check GCC so that we don't need to worry about the difference in command line args between GNU readelf and llvm-readelf/llvm-dwarfdump to validate the DWARF Version in the assembler feature detection script. GNU `as` only recently gained support for specifying -gdwarf-5, so when compiling with Clang but without Clang's integrated assembler (LLVM_IAS=1 is not set), explicitly add -Wa,-gdwarf-5 to DEBUG_CFLAGS. Disabled for now if CONFIG_DEBUG_INFO_BTF is set; pahole doesn't yet recognize the new additions to the DWARF debug info. Thanks to Sedat for the report. Link: http://www.dwarfstd.org/doc/DWARF5.pdf Reported-by: Sedat Dilek <sedat.dilek@gmail.com> Suggested-by: Arvind Sankar <nivedita@alum.mit.edu> Suggested-by: Caroline Tice <cmtice@google.com> Suggested-by: Fangrui Song <maskray@google.com> Suggested-by: Jakub Jelinek <jakub@redhat.com> Suggested-by: Masahiro Yamada <masahiroy@kernel.org> Suggested-by: Nathan Chancellor <natechancellor@gmail.com> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com> --- Makefile | 1 + include/asm-generic/vmlinux.lds.h | 7 ++++++- lib/Kconfig.debug | 18 ++++++++++++++++++ scripts/test_dwarf5_support.sh | 8 ++++++++ 4 files changed, 33 insertions(+), 1 deletion(-) create mode 100755 scripts/test_dwarf5_support.sh